0.9.8.34:
[sbcl.git] / src / code / fd-stream.lisp
index 0ef85b1..90621d9 100644 (file)
                             start
                             length)
       (cond ((not count)
-             (if (= errno sb!unix:ewouldblock)
+             (if #!-win32 (= errno sb!unix:ewouldblock) #!+win32 t #!-win32
                  (error "Write would have blocked, but SERVER told us to go.")
                  (simple-stream-perror "couldn't write to ~S" stream errno)))
             ((eql count length) ; Hot damn, it worked.
         (multiple-value-bind (count errno)
             (sb!unix:unix-write (fd-stream-fd stream) base start length)
           (cond ((not count)
-                 (if (= errno sb!unix:ewouldblock)
+                 (if #!-win32 (= errno sb!unix:ewouldblock) #!+win32 t #!-win32
                      (output-later stream base start end reuse-sap)
                      (simple-stream-perror "couldn't write to ~S"
                                            stream
         (end (or end (length (the vector thing)))))
     (declare (fixnum start end))
     (if (stringp thing)
-        (let ((last-newline (and (find #\newline (the simple-string thing)
-                                       :start start :end end)
-                                 ;; FIXME why do we need both calls?
-                                 ;; Is find faster forwards than
-                                 ;; position is backwards?
-                                 (position #\newline (the simple-string thing)
-                                           :from-end t
-                                           :start start
-                                           :end end))))
+        (let ((last-newline
+               (string-dispatch (simple-base-string
+                                 #!+sb-unicode
+                                 (simple-array character)
+                                 string)
+                   thing
+                 (and (find #\newline thing :start start :end end)
+                      ;; FIXME why do we need both calls?
+                      ;; Is find faster forwards than
+                      ;; position is backwards?
+                      (position #\newline thing
+                                :from-end t
+                                :start start
+                                :end end)))))
           (if (and (typep thing 'base-string)
                    (eq (fd-stream-external-format stream) :latin-1))
               (ecase (fd-stream-buffering stream)
                            (sb!sys:int-sap (+ (sb!sys:sap-int ibuf-sap) tail))
                            (- buflen tail))
       (cond ((null count)
-             (if (eql errno sb!unix:ewouldblock)
+             (if #!-win32 (eql errno sb!unix:ewouldblock) #!+win32 t #!-win32
                  (progn
                    (unless (sb!sys:wait-until-fd-usable
                             fd :input (fd-stream-timeout stream))
       (return-from fd-stream-resync
         (funcall (symbol-function (eighth entry)) stream)))))
 
+(defun get-fd-stream-character-sizer (stream)
+  (dolist (entry *external-formats*)
+    (when (member (fd-stream-external-format stream) (first entry))
+      (return-from get-fd-stream-character-sizer (ninth entry)))))
+
+(defun fd-stream-character-size (stream char)
+  (let ((sizer (get-fd-stream-character-sizer stream)))
+    (when sizer (funcall sizer char))))
+
+(defun fd-stream-string-size (stream string)
+  (let ((sizer (get-fd-stream-character-sizer stream)))
+    (when sizer
+      (loop for char across string summing (funcall sizer char)))))
+
 ;;; FIXME: OAOOM here vrt. *EXTERNAL-FORMAT-FUNCTIONS* in fd-stream.lisp
 (defmacro define-external-format (external-format size output-restart
                                   out-expr in-expr)
          (out-function (symbolicate "OUTPUT-BYTES/" name))
          (format (format nil "OUTPUT-CHAR-~A-~~A-BUFFERED" (string name)))
          (in-function (symbolicate "FD-STREAM-READ-N-CHARACTERS/" name))
-         (in-char-function (symbolicate "INPUT-CHAR/" name)))
+         (in-char-function (symbolicate "INPUT-CHAR/" name))
+         (size-function (symbolicate "BYTES-FOR-CHAR/" name)))
     `(progn
+      (defun ,size-function (byte)
+        (declare (ignore byte))
+        ,size)
       (defun ,out-function (stream string flush-p start end)
         (let ((start (or start 0))
               (end (or end (length string))))
                      (> (fd-stream-ibuf-tail stream)
                         (fd-stream-ibuf-head stream)))
             (file-position stream (file-position stream)))
-          (when (< end start)
-            (error ":END before :START!"))
+          (unless (<= 0 start end (length string))
+            (signal-bounding-indices-bad-error string start end))
           (do ()
               ((= end start))
             (setf (fd-stream-obuf-tail stream)
-                  (do* ((len (fd-stream-obuf-length stream))
-                        (sap (fd-stream-obuf-sap stream))
-                        (tail (fd-stream-obuf-tail stream)))
-                       ((or (= start end) (< (- len tail) 4)) tail)
-                    ,(if output-restart
-                         `(catch 'output-nothing
-                            (let* ((byte (aref string start))
-                                   (bits (char-code byte)))
-                              ,out-expr
-                              (incf tail ,size)))
-                         `(let* ((byte (aref string start))
-                                  (bits (char-code byte)))
-                             ,out-expr
-                             (incf tail ,size)))
-                    (incf start)))
+                  (string-dispatch (simple-base-string
+                                    #!+sb-unicode
+                                    (simple-array character)
+                                    string)
+                      string
+                    (let ((len (fd-stream-obuf-length stream))
+                          (sap (fd-stream-obuf-sap stream))
+                          (tail (fd-stream-obuf-tail stream)))
+                      (declare (type index tail)
+                               ;; STRING bounds have already been checked.
+                               (optimize (safety 0)))
+                      (loop
+                         (,@(if output-restart
+                                `(catch 'output-nothing)
+                                `(progn))
+                            (do* ()
+                                 ((or (= start end) (< (- len tail) 4)))
+                              (let* ((byte (aref string start))
+                                     (bits (char-code byte)))
+                                ,out-expr
+                                (incf tail ,size)
+                                (incf start)))
+                            ;; Exited from the loop normally
+                            (return tail))
+                         ;; Exited via CATCH. Skip the current character
+                         ;; and try the inner loop again.
+                         (incf start)))))
             (when (< start end)
               (flush-output-buffer stream)))
           (when flush-p
               (tail (fd-stream-obuf-tail stream)))
           ,out-expr))
       (defun ,in-function (stream buffer start requested eof-error-p
-                           &aux (total-copied 0))
+                           &aux (index start) (end (+ start requested)))
         (declare (type fd-stream stream))
-        (declare (type index start requested total-copied))
+        (declare (type index start requested index end))
         (declare (type (simple-array character (#.+ansi-stream-in-buffer-length+)) buffer))
         (let ((unread (fd-stream-unread stream)))
           (when unread
-            (setf (aref buffer start) unread)
+            (setf (aref buffer index) unread)
             (setf (fd-stream-unread stream) nil)
             (setf (fd-stream-listen stream) nil)
-            (incf total-copied)))
+            (incf index)))
         (do ()
             (nil)
           (let* ((head (fd-stream-ibuf-head stream))
                  (tail (fd-stream-ibuf-tail stream))
                  (sap (fd-stream-ibuf-sap stream)))
-            (declare (type index head tail))
+            (declare (type index head tail)
+                     (type system-area-pointer sap))
             ;; Copy data from stream buffer into user's buffer.
-            (do ()
-                ((or (= tail head) (= requested total-copied)))
+            (dotimes (i (min (truncate (- tail head) ,size)
+                             (- end index)))
+              (declare (optimize speed))
               (let* ((byte (sap-ref-8 sap head)))
-                (when (> ,size (- tail head))
-                  (return))
-                (setf (aref buffer (+ start total-copied)) ,in-expr)
-                (incf total-copied)
+                (setf (aref buffer index) ,in-expr)
+                (incf index)
                 (incf head ,size)))
             (setf (fd-stream-ibuf-head stream) head)
             ;; Maybe we need to refill the stream buffer.
-            (cond ( ;; If there were enough data in the stream buffer, we're done.
-                   (= total-copied requested)
-                   (return total-copied))
+            (cond ( ;; If there was enough data in the stream buffer, we're done.
+                   (= index end)
+                   (return (- index start)))
                   ( ;; If EOF, we're done in another way.
                    (null (catch 'eof-input-catcher (refill-buffer/fd stream)))
                    (if eof-error-p
                        (error 'end-of-file :stream stream)
-                       (return total-copied)))
+                       (return (- index start))))
                   ;; Otherwise we refilled the stream buffer, so fall
                   ;; through into another pass of the loop.
                   ))))
        (cons '(,external-format ,in-function ,in-char-function ,out-function
                ,@(mapcar #'(lambda (buffering)
                              (intern (format nil format (string buffering))))
-                         '(:none :line :full)))
+                         '(:none :line :full))
+               nil ; no resync-function
+               ,size-function)
         *external-formats*)))))
 
 (defmacro define-external-format/variable-width
          (format (format nil "OUTPUT-CHAR-~A-~~A-BUFFERED" (string name)))
          (in-function (symbolicate "FD-STREAM-READ-N-CHARACTERS/" name))
          (in-char-function (symbolicate "INPUT-CHAR/" name))
-         (resync-function (symbolicate "RESYNC/" name)))
+         (resync-function (symbolicate "RESYNC/" name))
+         (size-function (symbolicate "BYTES-FOR-CHAR/" name)))
     `(progn
+      (defun ,size-function (byte)
+        ,out-size-expr)
       (defun ,out-function (stream string flush-p start end)
         (let ((start (or start 0))
               (end (or end (length string))))
                      (> (fd-stream-ibuf-tail stream)
                         (fd-stream-ibuf-head stream)))
             (file-position stream (file-position stream)))
-          (when (< end start)
-            (error ":END before :START!"))
+          (unless (<= 0 start end (length string))
+            (signal-bounding-indices-bad-error string start end))
           (do ()
               ((= end start))
             (setf (fd-stream-obuf-tail stream)
-                  (do* ((len (fd-stream-obuf-length stream))
-                        (sap (fd-stream-obuf-sap stream))
-                        (tail (fd-stream-obuf-tail stream)))
-                       ((or (= start end) (< (- len tail) 4)) tail)
-                    ,(if output-restart
-                         `(catch 'output-nothing
-                            (let* ((byte (aref string start))
-                                   (bits (char-code byte))
-                                   (size ,out-size-expr))
-                              ,out-expr
-                              (incf tail size)))
-                         `(let* ((byte (aref string start))
-                                 (bits (char-code byte))
-                                 (size ,out-size-expr))
-                            ,out-expr
-                            (incf tail size)))
-                    (incf start)))
+                  (string-dispatch (simple-base-string
+                                    #!+sb-unicode
+                                    (simple-array character)
+                                    string)
+                      string
+                    (let ((len (fd-stream-obuf-length stream))
+                          (sap (fd-stream-obuf-sap stream))
+                          (tail (fd-stream-obuf-tail stream)))
+                      (declare (type index tail)
+                               ;; STRING bounds have already been checked.
+                               (optimize (safety 0)))
+                      (loop
+                         (,@(if output-restart
+                                `(catch 'output-nothing)
+                                `(progn))
+                            (do* ()
+                                 ((or (= start end) (< (- len tail) 4)))
+                              (let* ((byte (aref string start))
+                                     (bits (char-code byte))
+                                     (size ,out-size-expr))
+                                ,out-expr
+                                (incf tail size)
+                                (incf start)))
+                            ;; Exited from the loop normally
+                            (return tail))
+                         ;; Exited via CATCH. Skip the current character
+                         ;; and try the inner loop again.
+                         (incf start)))))
             (when (< start end)
               (flush-output-buffer stream)))
           (when flush-p
           (let* ((head (fd-stream-ibuf-head stream))
                  (tail (fd-stream-ibuf-tail stream))
                  (sap (fd-stream-ibuf-sap stream))
-                 (head-start head)
                  (decode-break-reason nil))
             (declare (type index head tail))
             ;; Copy data from stream buffer into user's buffer.
                         (incf head size))
                       nil))
               (setf (fd-stream-ibuf-head stream) head)
-              (when (and decode-break-reason
-                         (= head head-start))
+              (when decode-break-reason
+                ;; If we've already read some characters on when the invalid
+                ;; code sequence is detected, we return immediately. The
+                ;; handling of the error is deferred until the next call
+                ;; (where this check will be false). This allows establishing
+                ;; high-level handlers for decode errors (for example
+                ;; automatically resyncing in Lisp comments).
+                (when (plusp total-copied)
+                  (return-from ,in-function total-copied))
                 (when (stream-decoding-error-and-handle
                        stream decode-break-reason)
                   (if eof-error-p
                       (error 'end-of-file :stream stream)
                       (return-from ,in-function total-copied)))
                 (setf head (fd-stream-ibuf-head stream))
-                (setf tail (fd-stream-ibuf-tail stream)))
-              (when (plusp total-copied)
-                (return-from ,in-function total-copied)))
+                (setf tail (fd-stream-ibuf-tail stream))))
             (setf (fd-stream-ibuf-head stream) head)
             ;; Maybe we need to refill the stream buffer.
             (cond ( ;; If there were enough data in the stream buffer, we're done.
                ,@(mapcar #'(lambda (buffering)
                              (intern (format nil format (string buffering))))
                          '(:none :line :full))
-               ,resync-function)
+               ,resync-function
+               ,size-function)
         *external-formats*)))))
 
 ;;; Multiple names for the :ISO{,-}8859-* families are needed because on
        (if (zerop mode)
            nil
            (truncate size (fd-stream-element-size fd-stream)))))
-    ;; FIXME: I doubt this is correct in the presence of Unicode,
-    ;; since fd-stream FILE-POSITION is measured in bytes.
     (:file-string-length
      (etypecase arg1
-       (character 1)
-       (string (length arg1))))
+       (character (fd-stream-character-size fd-stream arg1))
+       (string (fd-stream-string-size fd-stream arg1))))
     (:file-position
      (fd-stream-file-position fd-stream arg1))))